|
2 | 2 | title: CPU and AVX |
3 | 3 | short_title: CPU |
4 | 4 | description: CPU and AVX support in ND4J/Deeplearning4j |
5 | | -layout: default |
| 5 | +layout: redirect |
| 6 | +redirectTo: https://deeplearning4j.konduit.ai/config/backends/cpu |
6 | 7 | --- |
7 | 8 |
|
8 | | -# ND4J - CPU (nd4j-native) AVX Configuration |
| 9 | +The documentation has moved to a new location. You will be automatically redirected. |
| 10 | + |
| 11 | +[Go to CPU and AVX without waiting for the redirect](https://deeplearning4j.konduit.ai/config/backends/cpu) |
9 | 12 |
|
10 | | -### What is AVX, and why does it matter? |
11 | | - |
12 | | -AVX (Advanced Vector Extensions) is a set of CPU instructions for accelerating numerical computations. See [Wikipedia](https://en.wikipedia.org/wiki/Advanced_Vector_Extensions) for more details. |
13 | | - |
14 | | -Note that AVX only applies to nd4j-native (CPU) backend for x86 devices, not GPUs and not ARM/PPC devices. |
15 | | - |
16 | | -Why AVX matters: performance. You want to use the version of ND4J compiled with the highest level of AVX supported by your system. |
17 | | - |
18 | | - |
19 | | -AVX support for different CPUs - summary: |
20 | | -* Most modern x86 CPUs: AVX2 is supported |
21 | | -* Some high-end server CPUs: AVX512 may be supported |
22 | | -* Old CPUs (pre 2012) and low power x86 (Atom, Celeron): No AVX support (usually) |
23 | | - |
24 | | -Note that CPUs supporting later versions of AVX include all earlier versions also. This means it's possible run a generic x86 or AVX2 binary on a system supporting AVX512. |
25 | | -However it is not possible to run binaries built for later versions (such as avx512) on a CPU that doesn't have support for those instructions. |
26 | | - |
27 | | -Note on current snapshots (and in future releases, after 1.0.0-beta5) you may get a warning as follows, if AVX is not configured optimally: |
28 | | -``` |
29 | | -o.n.l.c.n.CpuNDArrayFactory - *********************************** CPU Feature Check Warning *********************************** |
30 | | -o.n.l.c.n.CpuNDArrayFactory - Warning: Initializing ND4J with Generic x86 binary on a CPU with AVX/AVX2 support |
31 | | -o.n.l.c.n.CpuNDArrayFactory - Using ND4J with AVX/AVX2 will improve performance. See deeplearning4j.org/cpu for more details |
32 | | -o.n.l.c.n.CpuNDArrayFactory - Or set environment variable ND4J_IGNORE_AVX=true to suppress this warning |
33 | | -o.n.l.c.n.CpuNDArrayFactory - ************************************************************************************************ |
34 | | -``` |
35 | | - |
36 | | - |
37 | | -### Configuring AVX in ND4J/DL4J |
38 | | - |
39 | | -As noted earlier, for best performance you should use the version of ND4J that matches your CPU's supported AVX level. |
40 | | - |
41 | | -ND4J defaults configuration (when just including the nd4j-native or nd4j-native-platform dependencies without maven classifier configuration): |
42 | | -* 1.0.0-beta5 and earlier: "generic x86" (no AVX) is the default for nd4j/nd4j-platform dependencies |
43 | | -* Current snapshots and later versions of ND4J: AVX2 is the default |
44 | | - |
45 | | - |
46 | | -To configure AVX2 and AVX512, you need to specify a classifier for the appropriate architecture. |
47 | | - |
48 | | -The following binaries (nd4j-native classifiers) are provided for x86 architectures: |
49 | | -* Generic x86 (no AVX): `linux-x86_64`, `windows-x86_64`, `macosx-x86_64` |
50 | | -* AVX2: `linux-x86_64-avx2`, `windows-x86_64-avx2`, `macosx-x86_64-avx2` |
51 | | -* AVX512: `linux-x86_64-avx512` |
52 | | - |
53 | | - |
54 | | -**Example: Configuring AVX2 on Windows (Maven pom.xml)** |
55 | | -``` |
56 | | -<dependency> |
57 | | - <groupId>org.nd4j</groupId> |
58 | | - <artifactId>nd4j-native</artifactId> |
59 | | - <version>${nd4j.version}</version> |
60 | | -</dependency> |
61 | | -
|
62 | | -<dependency> |
63 | | - <groupId>org.nd4j</groupId> |
64 | | - <artifactId>nd4j-native</artifactId> |
65 | | - <version>${nd4j.version}</version> |
66 | | - <classifier>windows-x86_64-avx2</classifier> |
67 | | -</dependency> |
68 | | -``` |
69 | | - |
70 | | - |
71 | | -**Example: Configuring AVX512 on Linux (Maven pom.xml)** |
72 | | -``` |
73 | | -<dependency> |
74 | | - <groupId>org.nd4j</groupId> |
75 | | - <artifactId>nd4j-native</artifactId> |
76 | | - <version>${nd4j.version}</version> |
77 | | -</dependency> |
78 | | -
|
79 | | -<dependency> |
80 | | - <groupId>org.nd4j</groupId> |
81 | | - <artifactId>nd4j-native</artifactId> |
82 | | - <version>${nd4j.version}</version> |
83 | | - <classifier>linux-x86_64-avx512</classifier> |
84 | | -</dependency> |
85 | | -``` |
86 | | - |
87 | | -Note that you need *both* nd4j-native dependencies - with and without the classifier. |
| 13 | + |
0 commit comments